As a general rule, smoke and mirrors anywhere there are rough edges. In your case, if your demo does not work under less-than-ideal speech recognition conditions, I would not let them speak to the prototype. You could ask the audience for queries and speak them yourself.
Adding to this if voice recognition is just the first part of a process can you eliminate it? Can you show them an example of how to use it based on text? For example “type your command/query” instead of speaking it.
